GAME ENGAGEMENT PREDICTION¶

Importing Libraries¶

In [1]:
import sys,os
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
from scipy import stats
import warnings
warnings.filterwarnings('ignore')
from ydata_profiling import ProfileReport
from sklearn.preprocessing import StandardScaler,MinMaxScaler
from sklearn.preprocessing import LabelEncoder
from scipy.stats import skew
from skimpy import skim
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import (RandomForestClassifier,BaggingClassifier,AdaBoostClassifier,
                              GradientBoostingClassifier)
from xgboost import XGBClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import (accuracy_score,precision_score,recall_score,
                             confusion_matrix,classification_report,roc_auc_score,roc_curve)
from sklearn.model_selection import cross_val_score

Importing Dataset¶

In [2]:
df=pd.read_csv(r'C:\Users\VAIBHAV\Downloads\online_gaming_behavior_dataset.csv')

Data Exploration¶

In [3]:
# Looking at the first 4 Records of the dataset
df.head(4)
Out[3]:
PlayerID Age Gender Location GameGenre PlayTimeHours InGamePurchases GameDifficulty SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked EngagementLevel
0 9000 43 Male Other Strategy 16.271119 0 Medium 6 108 79 25 Medium
1 9001 29 Female USA Strategy 5.525961 0 Medium 5 144 11 10 Medium
2 9002 22 Female USA Sports 8.223755 0 Easy 16 142 35 41 High
3 9003 35 Male USA Action 5.265351 1 Easy 9 85 57 47 Medium
In [4]:
# Looking at last 3 record of the dataset
df.tail(3)
Out[4]:
PlayerID Age Gender Location GameGenre PlayTimeHours InGamePurchases GameDifficulty SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked EngagementLevel
40031 49031 15 Female USA RPG 0.240057 1 Easy 10 176 29 1 High
40032 49032 34 Male USA Sports 14.017818 1 Medium 3 128 70 10 Medium
40033 49033 19 Male USA Sports 10.083804 0 Easy 13 84 72 39 Medium
In [5]:
df.sample(2)
Out[5]:
PlayerID Age Gender Location GameGenre PlayTimeHours InGamePurchases GameDifficulty SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked EngagementLevel
38863 47863 41 Female USA Simulation 10.733632 0 Easy 16 152 58 32 High
25365 34365 29 Female USA Strategy 5.989988 0 Hard 2 25 23 0 Low
In [6]:
# shape of the dataset
df.shape
Out[6]:
(40034, 13)
In [7]:
# Number of rows and columns of dataset
print('Number of Rows:',df.shape[0])
print('Number of Columns:',df.shape[1])

Number of Rows: 40034
Number of Columns: 13
In [8]:
# A quick summary/overview of dataset 
skim(df)

╭──────────────────────────────────────────────── skimpy summary ─────────────────────────────────────────────────╮
│          Data Summary                Data Types                                                                 │
│ ┏━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┓ ┏━━━━━━━━━━━━━┳━━━━━━━┓                                                          │
│ ┃ dataframe         ┃ Values ┃ ┃ Column Type ┃ Count ┃                                                          │
│ ┡━━━━━━━━━━━━━━━━━━━╇━━━━━━━━┩ ┡━━━━━━━━━━━━━╇━━━━━━━┩                                                          │
│ │ Number of rows    │ 40034  │ │ int32       │ 7     │                                                          │
│ │ Number of columns │ 13     │ │ string      │ 5     │                                                          │
│ └───────────────────┴────────┘ │ float64     │ 1     │                                                          │
│                                └─────────────┴───────┘                                                          │
│                                                     number                                                      │
│ ┏━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━┳━━━━━━━┳━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━━┓  │
│ ┃ column_name           ┃ NA  ┃ NA %  ┃ mean   ┃ sd     ┃ p0        ┃ p25   ┃ p50   ┃ p75   ┃ p100  ┃ hist   ┃  │
│ ┡━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━╇━━━━━━━╇━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━╇━━━━━━━╇━━━━━━━╇━━━━━━━╇━━━━━━━━┩  │
│ │ PlayerID              │   0 │     0 │  29020 │  11560 │      9000 │ 19010 │ 29020 │ 39020 │ 49030 │ ▇▇▇▇▇▇ │  │
│ │ Age                   │   0 │     0 │  31.99 │  10.04 │        15 │    23 │    32 │    41 │    49 │ ▇▇▇▇▇▇ │  │
│ │ PlayTimeHours         │   0 │     0 │  12.02 │  6.915 │ 0.0001147 │ 6.068 │ 12.01 │ 17.96 │    24 │ ▇▇▇▇▇▇ │  │
│ │ InGamePurchases       │   0 │     0 │ 0.2009 │ 0.4006 │         0 │     0 │     0 │     0 │     1 │ ▇    ▂ │  │
│ │ SessionsPerWeek       │   0 │     0 │  9.472 │  5.764 │         0 │     4 │     9 │    14 │    19 │ ▇▆▆▆▆▇ │  │
│ │ AvgSessionDurationMin │   0 │     0 │  94.79 │  49.01 │        10 │    52 │    95 │   137 │   179 │ ▇▇▇▇▇▇ │  │
│ │ utes                  │     │       │        │        │           │       │       │       │       │        │  │
│ │ PlayerLevel           │   0 │     0 │  49.66 │  28.59 │         1 │    25 │    49 │    74 │    99 │ ▇▇▇▇▇▇ │  │
│ │ AchievementsUnlocked  │   0 │     0 │  24.53 │  14.43 │         0 │    12 │    25 │    37 │    49 │ ▇▇▇▇▇▇ │  │
│ └───────────────────────┴─────┴───────┴────────┴────────┴───────────┴───────┴───────┴───────┴───────┴────────┘  │
│                                                     string                                                      │
│ ┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━┓  │
│ ┃ column_name                     ┃ NA     ┃ NA %       ┃ words per row              ┃ total words           ┃  │
│ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━┩  │
│ │ Gender                          │      0 │          0 │                          1 │                 40034 │  │
│ │ Location                        │      0 │          0 │                          1 │                 40034 │  │
│ │ GameGenre                       │      0 │          0 │                          1 │                 40034 │  │
│ │ GameDifficulty                  │      0 │          0 │                          1 │                 40034 │  │
│ │ EngagementLevel                 │      0 │          0 │                          1 │                 40034 │  │
│ └─────────────────────────────────┴────────┴────────────┴────────────────────────────┴───────────────────────┘  │
╰────────────────────────────────────────────────────── End ──────────────────────────────────────────────────────╯
In [9]:
# looking at column names
df.columns
Out[9]:
Index(['PlayerID', 'Age', 'Gender', 'Location', 'GameGenre', 'PlayTimeHours',
       'InGamePurchases', 'GameDifficulty', 'SessionsPerWeek',
       'AvgSessionDurationMinutes', 'PlayerLevel', 'AchievementsUnlocked',
       'EngagementLevel'],
      dtype='object')
In [10]:
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 40034 entries, 0 to 40033
Data columns (total 13 columns):
 #   Column                     Non-Null Count  Dtype  
---  ------                     --------------  -----  
 0   PlayerID                   40034 non-null  int64  
 1   Age                        40034 non-null  int64  
 2   Gender                     40034 non-null  object 
 3   Location                   40034 non-null  object 
 4   GameGenre                  40034 non-null  object 
 5   PlayTimeHours              40034 non-null  float64
 6   InGamePurchases            40034 non-null  int64  
 7   GameDifficulty             40034 non-null  object 
 8   SessionsPerWeek            40034 non-null  int64  
 9   AvgSessionDurationMinutes  40034 non-null  int64  
 10  PlayerLevel                40034 non-null  int64  
 11  AchievementsUnlocked       40034 non-null  int64  
 12  EngagementLevel            40034 non-null  object 
dtypes: float64(1), int64(7), object(5)
memory usage: 4.0+ MB
In [11]:
# Checking for null values
df.isnull().sum()
Out[11]:
PlayerID                     0
Age                          0
Gender                       0
Location                     0
GameGenre                    0
PlayTimeHours                0
InGamePurchases              0
GameDifficulty               0
SessionsPerWeek              0
AvgSessionDurationMinutes    0
PlayerLevel                  0
AchievementsUnlocked         0
EngagementLevel              0
dtype: int64
There are no null values in the dataset¶
In [12]:
# checking for duplicates
df.duplicated().sum()
Out[12]:
0
In [13]:
# Dropping column PlayerId as it is not significant for model building
df.drop('PlayerID',axis=1,inplace=True)
In [14]:
# Separating numerical and categorical columns
df_num=df.select_dtypes(include='number')
df_cat=df.select_dtypes(include='object')
In [15]:
# unique values of categorical column
for i in df_cat.columns:
    print(i,df[i].unique())
    print('**********************************')

Gender ['Male' 'Female']
**********************************
Location ['Other' 'USA' 'Europe' 'Asia']
**********************************
GameGenre ['Strategy' 'Sports' 'Action' 'RPG' 'Simulation']
**********************************
GameDifficulty ['Medium' 'Easy' 'Hard']
**********************************
EngagementLevel ['Medium' 'High' 'Low']
**********************************
In [16]:
df['EngagementLevel'].value_counts()
Out[16]:
EngagementLevel
Medium    19374
High      10336
Low       10324
Name: count, dtype: int64
In [17]:
# unique values of numerical column
for i in df_num.columns:
    print(i,df[i].unique())
    print('**********************************')

Age [43 29 22 35 33 37 25 38 17 36 16 44 47 26 39 41 42 30 21 23 32 18 28 40
 34 49 31 20 48 24 45 15 19 27 46]
**********************************
PlayTimeHours [16.27111876  5.52596138  8.22375524 ...  0.24005688 14.01781798
 10.08380358]
**********************************
InGamePurchases [0 1]
**********************************
SessionsPerWeek [ 6  5 16  9  2  1 10 13  8  0  3  4 18 14 12 19 17 11 15  7]
**********************************
AvgSessionDurationMinutes [108 144 142  85 131  81  50  48 101  95 124  18 156  41 154 135  56 177
 159 120 117 161  82 118  57 155  67  70 127 166  28  61 112 116  43 151
  68 115 149  65 176  31  77  94 146  45 134 171  46 139  36 148 153 138
 140  16  49  30 167  79 111 109 106  29  83  14  34 102 122 147 107  73
  12  99  32 126  84  55  93  40  33  91  24  72  96 121  60 145  51  69
  75  17  62 157  78  21  42 103  52 132 105  27  19  39  92 179 164  90
 110 133  98  89  80  47  20  10 152 128  71 175  88 125  74 168  76 169
 172 170 150  23  87 160  26  63 174 165  66 104  44 158 162  59 178  86
  64 114 173 113 163  35 123  53 119  58 136  37  54  97 130  25 100 129
 143  38  13  11  15  22 141 137]
**********************************
PlayerLevel [79 11 35 57 95 74 13 27 23 99 14 62 52 33 98 58 77 21 34 36 81 40  1 71
 59  2 85 50 48 24 26 93 76 88 10 45 73 51 31 22 19 64  6  8 20 46 38 80
  3 72 29 12 91  9 15  4 89 78 61 54 65 53 56 75  5 42 69 49 92 96 82 87
 60 39 66 55 32  7 83 67 16 84 68 94 44 30 47 25 90 17 63 28 97 86 37 43
 70 41 18]
**********************************
AchievementsUnlocked [25 10 41 47 37 22  2 23 36 12 31 32  1 24  4  9 30 33  5 19  7 26 18 11
 48 42 17  0 44 29 35 39 34 46 43 49 38 13 21 16 40 15 28 14 45 20 27  3
  8  6]
**********************************
In [18]:
# Statistical summary of numerical columns
df_num.describe()
Out[18]:
Age PlayTimeHours InGamePurchases SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked
count 40034.000000 40034.000000 40034.000000 40034.000000 40034.000000 40034.000000 40034.000000
mean 31.992531 12.024365 0.200854 9.471774 94.792252 49.655568 24.526477
std 10.043227 6.914638 0.400644 5.763667 49.011375 28.588379 14.430726
min 15.000000 0.000115 0.000000 0.000000 10.000000 1.000000 0.000000
25% 23.000000 6.067501 0.000000 4.000000 52.000000 25.000000 12.000000
50% 32.000000 12.008002 0.000000 9.000000 95.000000 49.000000 25.000000
75% 41.000000 17.963831 0.000000 14.000000 137.000000 74.000000 37.000000
max 49.000000 23.999592 1.000000 19.000000 179.000000 99.000000 49.000000
In [19]:
df_cat.describe()
Out[19]:
Gender Location GameGenre GameDifficulty EngagementLevel
count 40034 40034 40034 40034 40034
unique 2 4 5 3 3
top Male USA Sports Easy Medium
freq 23959 16000 8048 20015 19374

Pandas Profiling¶

In [20]:
# This will give the detailed summary of the dataset
from ydata_profiling import ProfileReport
profile=ProfileReport(df,title='myfile')
profile.to_notebook_iframe()

Summarize dataset:   0%|          | 0/5 [00:00<?, ?it/s]
Generate report structure:   0%|          | 0/1 [00:00<?, ?it/s]
Render HTML:   0%|          | 0/1 [00:00<?, ?it/s]
In [21]:
# data extraction using boolean indexing
# looking at only male gamers from asia
df[(df['Gender']=='Male') & (df['Location']=='Asia')].head(3)
Out[21]:
Age Gender Location GameGenre PlayTimeHours InGamePurchases GameDifficulty SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked EngagementLevel
16 35 Male Asia Strategy 17.887898 0 Easy 16 154 62 4 High
45 20 Male Asia RPG 16.585329 0 Medium 19 65 34 46 Medium
71 15 Male Asia Strategy 18.850359 0 Medium 7 111 45 2 Medium
In [22]:
# df.style.background_gradient(cmap='viridis')

Data Visualization¶

In [23]:
# plt.style.use('dark_background')
sns.catplot(df,x='EngagementLevel',kind='count',palette='dark')
plt.title('Distribution of Engagement Level')
plt.xlabel('Engagement Level')
plt.ylabel('Count of players')
Out[23]:
Text(-12.430555555555555, 0.5, 'Count of players')
No description has been provided for this image
In [24]:
# for i in df_cat.columns:
#     plt.figure()
#     sns.countplot(df,y=i,palette='colorblind',order=df[i].value_counts().index)
#     plt.title(f'count of {i}')
#     plt.show()
In [25]:
fig,axes = plt.subplots(2,2,figsize=(11,11))

ax=sns.countplot(ax=axes[0, 0],x='Gender',data=df,palette='colorblind',order=df['Gender'].value_counts().index)
axes[0, 0].set_title('Distribution of Gender')

ax=sns.countplot(ax=axes[0, 1],x='Location',data=df,palette='colorblind',order=df['Location'].value_counts().index)
axes[0, 1].set_title('Distribution of Location')

ax=sns.countplot(ax=axes[1, 0],x='GameGenre',data=df,palette='colorblind',order=df['GameGenre'].value_counts().index)
axes[1, 0].set_title('Distribution of GameGenre')

ax=sns.countplot(ax=axes[1, 1],x='GameDifficulty',data=df,palette='colorblind',order=df['GameDifficulty'].value_counts().index)
axes[1, 1].set_title('Distribution of GameDifficulty')
plt.show()
No description has been provided for this image
In [26]:
fig,axes = plt.subplots(2, 2, figsize=(12,12))

# Gender grouped by EngagementLevel
ax=sns.countplot(ax=axes[0, 0], x='Gender', hue='EngagementLevel',data=df,palette='mako')
axes[0, 0].set_title('Gender grouped by EngagementLevel')

# Location grouped by EngagementLevel
ax=sns.countplot(ax=axes[0, 1], x='Location', hue='EngagementLevel',data=df,palette='mako')
axes[0, 1].set_title('Location grouped by EngagementLevel')

# GameGenre grouped by EngagementLevel
ax=sns.countplot(ax=axes[1, 0], x='GameGenre', hue='EngagementLevel',data=df,palette='mako')
axes[1, 0].set_title('GameGenre grouped by EngagementLevel')

# GameDifficulty grouped by EngagementLevel
ax=sns.countplot(ax=axes[1, 1], x='GameDifficulty',hue='EngagementLevel',data=df,palette='mako')
axes[1, 1].set_title('GameDifficulty grouped by EngagementLevel')
plt.show()
No description has been provided for this image
In [27]:
sns.catplot(df,x='Location',kind='count',hue='Gender',palette='bright')
Out[27]:
<seaborn.axisgrid.FacetGrid at 0x1e599dca390>
No description has been provided for this image
In [28]:
# pie chart
plt.figure(figsize=(5,6))
df['EngagementLevel'].value_counts().plot(kind='pie',textprops={'color':'black'},autopct='%.2f%%',cmap='cool')
plt.title('Distribution of EngagementLevel')
plt.legend(['Medium','High','Low'])
plt.show()
No description has been provided for this image
In [29]:
# distribution of age
sns.histplot(df['Age'],color='#098968',edgecolor='#00FFBD')
plt.ylabel('Count of Players ')
plt.title('Age Distribution')
Out[29]:
Text(0.5, 1.0, 'Age Distribution')
No description has been provided for this image
In [30]:
# distibution of playtime hours
sns.histplot(df['PlayTimeHours'],color='#098968',edgecolor='#00FFBD')
plt.ylabel('Count of Players ')
Out[30]:
Text(0, 0.5, 'Count of Players ')
No description has been provided for this image
In [31]:
df.head()
Out[31]:
Age Gender Location GameGenre PlayTimeHours InGamePurchases GameDifficulty SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked EngagementLevel
0 43 Male Other Strategy 16.271119 0 Medium 6 108 79 25 Medium
1 29 Female USA Strategy 5.525961 0 Medium 5 144 11 10 Medium
2 22 Female USA Sports 8.223755 0 Easy 16 142 35 41 High
3 35 Male USA Action 5.265351 1 Easy 9 85 57 47 Medium
4 33 Male Europe Action 15.531945 0 Medium 2 131 95 37 Medium
In [32]:
plt.style.use('dark_background')
plt.figure(figsize=(10, 6))
sns.violinplot( data=df,x='EngagementLevel', y='SessionsPerWeek',hue='EngagementLevel',palette='muted')
plt.title('SessionsPerWeek based on EngagementLevel')  #color='#098968',
Out[32]:
Text(0.5, 1.0, 'SessionsPerWeek based on EngagementLevel')
No description has been provided for this image
In [33]:
# distribution plot
for i in df_num.columns:
    sns.displot(df[i],kind='kde')
    plt.title(f'Distribution of {i}')
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
In [34]:
# boxplot
for i in df_num.columns:
    plt.figure()
    sns.boxplot(df[i])
    plt.title(f'Box plot for {i}')
    plt.show()
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image
No description has been provided for this image

Data Preprocessing¶

In [35]:
# Encoding for categorical columns
# label encoding for target columns Engagement Level and gender
from sklearn.preprocessing import LabelEncoder
le=LabelEncoder()
df['EngagementLevel']=le.fit_transform(df['EngagementLevel'])
df['Gender']=le.fit_transform(df['Gender'])
In [36]:
df_cat.columns
Out[36]:
Index(['Gender', 'Location', 'GameGenre', 'GameDifficulty', 'EngagementLevel'], dtype='object')
In [37]:
# one hot encoding for all other categorical columns
df=pd.get_dummies(data=df,columns=['Location','GameGenre','GameDifficulty'],dtype='int')
In [38]:
# display all columns
pd.set_option('display.max_columns',None)
In [39]:
# dropping last columns of the one hot coded feature to avoid multicollinearity
df.drop(['Location_USA','GameGenre_Strategy','GameDifficulty_Medium'],inplace=True,axis=1)
In [40]:
# or
# Df = pd.get_dummies(df,columns=["Location","GameGenre","GameDifficulty"],drop_first=True)
In [41]:
df.head(3)
Out[41]:
Age Gender PlayTimeHours InGamePurchases SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked EngagementLevel Location_Asia Location_Europe Location_Other GameGenre_Action GameGenre_RPG GameGenre_Simulation GameGenre_Sports GameDifficulty_Easy GameDifficulty_Hard
0 43 1 16.271119 0 6 108 79 25 2 0 0 1 0 0 0 0 0 0
1 29 0 5.525961 0 5 144 11 10 2 0 0 0 0 0 0 0 0 0
2 22 0 8.223755 0 16 142 35 41 0 0 0 0 0 0 0 1 1 0
In [42]:
# # Heatmap
# plt.figure(figsize=(12,10))
# plt.style.use('dark_background')
# sns.heatmap(df.corr(),annot=True,fmt='.2f',linecolor='white',cmap='Greens')

Seperating x and y¶

In [43]:
# Here we will seperate x (independent features) and y (dependent feature) also called the target variable
In [44]:
x=df.drop('EngagementLevel',axis=1)
y=df['EngagementLevel']
In [45]:
x.head(3)
Out[45]:
Age Gender PlayTimeHours InGamePurchases SessionsPerWeek AvgSessionDurationMinutes PlayerLevel AchievementsUnlocked Location_Asia Location_Europe Location_Other GameGenre_Action GameGenre_RPG GameGenre_Simulation GameGenre_Sports GameDifficulty_Easy GameDifficulty_Hard
0 43 1 16.271119 0 6 108 79 25 0 0 1 0 0 0 0 0 0
1 29 0 5.525961 0 5 144 11 10 0 0 0 0 0 0 0 0 0
2 22 0 8.223755 0 16 142 35 41 0 0 0 0 0 0 1 1 0
In [46]:
y.head(3)
Out[46]:
0    2
1    2
2    0
Name: EngagementLevel, dtype: int32

Train Test Split¶

In [47]:
from sklearn.model_selection import train_test_split
x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.2,random_state=101)

Model Building¶

Logistic Regression¶

In [48]:
from sklearn.linear_model import LogisticRegression
LR=LogisticRegression()
LR.fit(x_train,y_train)
Out[48]:
LogisticRegression()
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
LogisticRegression()
In [49]:
y_pred_tr=LR.predict(x_train)
y_pred_tst=LR.predict(x_test)

## Testing and Training Accuracy
from sklearn.metrics import accuracy_score
print('Accuracy score on Training Set (Linear Regression):',accuracy_score(y_train,y_pred_tr)*100)
print('Accuracy score on Testing Set (Linear Regression):' ,accuracy_score(y_test,y_pred_tst)*100)

Accuracy score on Training Set (Linear Regression): 72.4232678677366
Accuracy score on Testing Set (Linear Regression): 72.73635568877232
In [50]:
# Precision and Recall score
from sklearn.metrics import precision_score,recall_score
print('Precision Score:',precision_score(y_test,y_pred_tst,average='micro'))
print('Recall Score:',recall_score(y_test,y_pred_tst,average='micro'))

Precision Score: 0.7273635568877233
Recall Score: 0.7273635568877233
In [51]:
# classificatio Report
print(classification_report(y_test,y_pred_tst))

              precision    recall  f1-score   support

           0       0.73      0.68      0.70      2074
           1       0.80      0.62      0.70      2072
           2       0.70      0.82      0.75      3861

    accuracy                           0.73      8007
   macro avg       0.74      0.70      0.72      8007
weighted avg       0.73      0.73      0.73      8007
In [52]:
# Confusion matrix
from sklearn.metrics import ConfusionMatrixDisplay
cm=confusion_matrix(y_test,y_pred_tst)
In [53]:
cm_display=ConfusionMatrixDisplay(cm,display_labels=[0,1,2])
cm_display.plot()
plt.show()
No description has been provided for this image

Prediction On Scaled Data¶

In [54]:
from sklearn.preprocessing import StandardScaler
ss=StandardScaler()
x_train_s=ss.fit_transform(x_train)
x_test_s=ss.transform(x_test)
In [55]:
LR=LogisticRegression()
LR.fit(x_train_s,y_train)
Out[55]:
LogisticRegression()
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
LogisticRegression()
In [56]:
y_pred_tr=LR.predict(x_train_s)
y_pred_tst=LR.predict(x_test_s)
In [57]:
print('Accuracy score on Training set (Scaled data):',accuracy_score(y_train,y_pred_tr)*100)
print('Accuracy score on Testing set (Scaled data) :',accuracy_score(y_test,y_pred_tst)*100)

Accuracy score on Training set (Scaled data): 82.27745339869486
Accuracy score on Testing set (Scaled data) : 82.71512426626701
In [58]:
# classificatio Report
print(classification_report(y_test,y_pred_tst))

              precision    recall  f1-score   support

           0       0.90      0.84      0.87      2074
           1       0.80      0.70      0.75      2072
           2       0.80      0.89      0.85      3861

    accuracy                           0.83      8007
   macro avg       0.84      0.81      0.82      8007
weighted avg       0.83      0.83      0.83      8007
In [59]:
# Confusion matrix
from sklearn.metrics import ConfusionMatrixDisplay
cm=confusion_matrix(y_test,y_pred_tst)
cm_display=ConfusionMatrixDisplay(cm,display_labels=[0,1,2])
cm_display.plot()
plt.show()
No description has been provided for this image

Decision Tree¶

In [60]:
from sklearn.tree import DecisionTreeClassifier
DT=DecisionTreeClassifier(max_depth=26,min_samples_split=6,min_samples_leaf=4)
DT.fit(x_train,y_train)
Out[60]:
DecisionTreeClassifier(max_depth=26, min_samples_leaf=4, min_samples_split=6)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
DecisionTreeClassifier(max_depth=26, min_samples_leaf=4, min_samples_split=6)
In [61]:
y_pred_tr=DT.predict(x_train)
y_pred_tst=DT.predict(x_test)
In [62]:
print('Accuracy score on Training set (Decision Tree ) :',accuracy_score(y_train,y_pred_tr)*100)
print('Accuracy score on Testing set (Decision Tree )  :',accuracy_score(y_test,y_pred_tst)*100)

Accuracy score on Training set (Decision Tree ) : 93.47737846192275
Accuracy score on Testing set (Decision Tree )  : 86.98638691145248
In [63]:
DT.get_depth()
Out[63]:
26
In [64]:
# Confusion matrix
from sklearn.metrics import ConfusionMatrixDisplay
cm=confusion_matrix(y_test,y_pred_tst)
cm_display=ConfusionMatrixDisplay(cm,display_labels=[0,1,2])
cm_display.plot()
plt.show()
No description has been provided for this image

Random Forest¶

In [65]:
RF=RandomForestClassifier(n_estimators=122,min_samples_split=3,min_samples_leaf=2)  
RF.fit(x_train,y_train)
Out[65]:
RandomForestClassifier(min_samples_leaf=2, min_samples_split=3,
                       n_estimators=122)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
RandomForestClassifier(min_samples_leaf=2, min_samples_split=3,
                       n_estimators=122)
In [66]:
y_pred_tr=RF.predict(x_train)
y_pred_tst=RF.predict(x_test)
In [67]:
print('Accuracy score on Training set (Random Forest) :',accuracy_score(y_train,y_pred_tr)*100)
print('Accuracy score on Testing set (Random Forest)  :',accuracy_score(y_test,y_pred_tst)*100)

Accuracy score on Training set (Random Forest) : 94.60767477440909
Accuracy score on Testing set (Random Forest)  : 90.9204446109654
In [68]:
# Confusion matrix
from sklearn.metrics import ConfusionMatrixDisplay
cm=confusion_matrix(y_test,y_pred_tst)
cm_display=ConfusionMatrixDisplay(cm,display_labels=[0,1,2])
cm_display.plot()
plt.show()
No description has been provided for this image
In [69]:
# classificatio Report
print(classification_report(y_test,y_pred_tst))

              precision    recall  f1-score   support

           0       0.92      0.86      0.89      2074
           1       0.92      0.87      0.89      2072
           2       0.90      0.95      0.93      3861

    accuracy                           0.91      8007
   macro avg       0.91      0.90      0.90      8007
weighted avg       0.91      0.91      0.91      8007
In [70]:
# Cross validation score
from sklearn.model_selection import cross_val_score
print('For Training data:',cross_val_score(RF,x_train,y_train,cv=5,scoring='accuracy'))

For Training data: [0.90649391 0.90883547 0.89992194 0.90226386 0.91178767]

Gradient boost¶

In [71]:
GB=GradientBoostingClassifier()
GB.fit(x_train,y_train)
Out[71]:
GradientBoostingClassifier()
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GradientBoostingClassifier()
In [72]:
y_pred_tr=GB.predict(x_train)
y_pred_tst=GB.predict(x_test)
In [73]:
print('Accuracy on Training set (Gradient Boosting):',accuracy_score(y_train,y_pred_tr)*100)
print('Accuracy on Testing set (Gradient Boosting):' ,accuracy_score(y_test,y_pred_tst)*100)

Accuracy on Training set (Gradient Boosting): 91.34168045711431
Accuracy on Testing set (Gradient Boosting): 91.13275883601848
In [74]:
# Confusion matrix
from sklearn.metrics import ConfusionMatrixDisplay
cm=confusion_matrix(y_test,y_pred_tst)
cm_display=ConfusionMatrixDisplay(cm,display_labels=[0,1,2])
cm_display.plot()
plt.show()
No description has been provided for this image

Bagging¶

In [75]:
BG=BaggingClassifier(estimator=RandomForestClassifier(),n_estimators=12)
BG.fit(x_train,y_train)
Out[75]:
BaggingClassifier(estimator=RandomForestClassifier(), n_estimators=12)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
BaggingClassifier(estimator=RandomForestClassifier(), n_estimators=12)
RandomForestClassifier()
RandomForestClassifier()
In [76]:
y_pred_tr=BG.predict(x_train)
y_pred_test=BG.predict(x_test)
In [77]:
print('Model performance for Training set')
print("- Accuracy: {:.4f}".format(accuracy_score(y_train,y_pred_tr)*100))

print('----------------------------------')    

print('Model performance for Test set')
print("- Accuracy: {:.4f}".format(accuracy_score(y_test,y_pred_tst)*100))
    

Model performance for Training set
- Accuracy: 96.4093
----------------------------------
Model performance for Test set
- Accuracy: 91.1328

Support Vector machine¶

In [78]:
from sklearn.svm import SVC
svn=SVC()
svn.fit(x_train,y_train)
Out[78]:
SVC()
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
SVC()
In [79]:
y_pred_tr=svn.predict(x_train)
y_pred_tst=svn.predict(x_test)
In [80]:
print('Accuracy on Training set (SVM):',accuracy_score(y_train,y_pred_tr)*100)
print('Accuracy on Testing set (SVM):' ,accuracy_score(y_test,y_pred_tst)*100)

Accuracy on Training set (SVM): 90.09897898648016
Accuracy on Testing set (SVM): 89.88385162982391

Roc Curve¶

In [81]:
# multi-class classification
from sklearn.datasets import make_classification
from sklearn.multiclass import OneVsRestClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import roc_curve
from sklearn.metrics import roc_auc_score
from sklearn.ensemble import RandomForestClassifier

# generate 2 class dataset
x, y = make_classification(n_samples=1000, n_classes=3, n_features=20, n_informative=3, random_state=42)

# split into train/test sets
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=42)

# fit model
clf = OneVsRestClassifier(RandomForestClassifier())
clf.fit(x_train, y_train)
pred = clf.predict(x_test)
pred_prob = clf.predict_proba(x_test)

# roc curve for classes
fpr = {}
tpr = {}
thresh ={}

n_class = 3

for i in range(n_class):    
    fpr[i], tpr[i], thresh[i] = roc_curve(y_test,pred_prob[:,i], pos_label=i)

# plotting    
plt.plot(fpr[0], tpr[0], linestyle='--',color='orange', label='Class 0 vs Rest')
plt.plot(fpr[1], tpr[1], linestyle='--',color='green', label='Class 1 vs Rest')
plt.plot(fpr[2], tpr[2], linestyle='--',color='blue', label='Class 2 vs Rest')
plt.title('Multiclass ROC curve')
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive rate')
plt.legend(loc='best')
Out[81]:
<matplotlib.legend.Legend at 0x1e597a9a690>
No description has been provided for this image

Actual vs Predicted¶

In [84]:
x=df.drop('EngagementLevel',axis=1)
y=df['EngagementLevel']
In [85]:
data=x.head(10)
In [86]:
y_pred=RF.predict(data)
y_pred
Out[86]:
array([2, 2, 0, 2, 2, 1, 1, 2, 2, 0])
In [87]:
data['Actual']=y.head(10)
data['Predicted']=y_pred
data[['Actual','Predicted']]
Out[87]:
Actual Predicted
0 2 2
1 2 2
2 0 0
3 2 2
4 2 2
5 1 1
6 1 1
7 2 2
8 2 2
9 0 0

Accuracy table for different models¶

In [82]:
df_accuracy=pd.DataFrame(index=list(range(1,8)))
df_accuracy['Model Name']=['Logistic Regression','Scaled Data','Decision Tree','Random Forest','GradientBoost','Bagging',\
                            'SVM']
df_accuracy['Accuracy Score(Training)']=np.round([72.4232678677366,82.27745339869486,93.47737846192275,94.610797139913,\
                                           91.34168045711431,91.34168045711431,90.09897898648016],2)
df_accuracy['Accuracy Score(Testing)']=np.round([72.73635568877232,82.71512426626701,86.99887598351442,90.99537904333708,91.13275883601848,\
                                           91.1328,89.8838516298239],2)
df_accuracy['Error']=100-df_accuracy['Accuracy Score(Testing)']

df_accuracy
Out[82]:
Model Name Accuracy Score(Training) Accuracy Score(Testing) Error
1 Logistic Regression 72.42 72.74 27.26
2 Scaled Data 82.28 82.72 17.28
3 Decision Tree 93.48 87.00 13.00
4 Random Forest 94.61 91.00 9.00
5 GradientBoost 91.34 91.13 8.87
6 Bagging 91.34 91.13 8.87
7 SVM 90.10 89.88 10.12
In [83]:
plt.figure(figsize=(20,6))
sns.barplot(x=df_accuracy['Model Name'],y=df_accuracy['Accuracy Score(Testing)'],palette='bright')
Out[83]:
<Axes: xlabel='Model Name', ylabel='Accuracy Score(Testing)'>
No description has been provided for this image
In [ ]:
In [2]:
# LR=LogisticRegression()
# DT=DecisionTreeClassifier(max_depth=26,min_samples_split=6,min_samples_leaf=4)
# RF=RandomForestClassifier(n_estimators=122,min_samples_split=3,min_samples_leaf=2)
# BG=BaggingClassifier()
# ADB=AdaBoostClassifier()
# GB=GradientBoostingClassifier()
# XGB=XGBClassifier()
# KNN=KNeighborsClassifier()
# NB=GaussianNB()
# SVN=SVC()
In [3]:
# lst=[LR,DT,RF,BG,ADB,GB,XGB,KNN,NB,SVN]
In [1]:
# for i in lst:
#     print('Classifier :',i)
#     i.fit(x_train,y_train)
#     y_pred_tr=i.predict(x_train)
#     y_pred_tst=i.predict(x_test)
#     print('Accuracy Score on Training Data',accuracy_score(y_train,y_pred_tr)*100)
#     print('Accuracy Score on Testing Data',accuracy_score(y_test,y_pred_tst)*100)
#     print('********************************************************')
#     print('********************************************************')
In [ ]:
In [ ]:
In [ ]: